首页|Researchers Submit Patent Application, 'Generating And/Or Using Training Instances That Include Previously Captured Robot Vision Data And Drivability Labels', for Approval (USPTO 20230401419)
Researchers Submit Patent Application, 'Generating And/Or Using Training Instances That Include Previously Captured Robot Vision Data And Drivability Labels', for Approval (USPTO 20230401419)
扫码查看
点击上方二维码区域，可以放大扫码查看
原文链接
NETL
NSTL
News editors obtained the following quote from the background information supplied by the inventors:“Limitations on processing of raw sensor data that is captured when a robot is traversing relatively unfamiliarterrains can cause the robot to experience delays in decision making- particularly with respect to navigation.Such limitations can result from an inability of the robot to reliably perform real-time analysis of forthcomingterrains that are observable along a current route. Although a robot may be able to reliably estimatedrivability of a surface that is directly adjacent to the robot, the robot may not be able to, with the sameor similar accuracy, estimate drivability of surfaces that are less proximate to the robot-at least not withoutfurther approaching such surfaces. For example, when a robot is traversing a room that the robot has notpreviously driven through, the robot may employ a heuristic approach to identify surfaces upon which therobot can readily traverse. Such an approach can rely on near-field vision data and can limit the abilityof the robot to operate in unfamiliar and/or dynamically changing locations with the same efficiency andaccuracy as the robot operates in more familiar terrains.”
NETL
NSTL
